Likeable Lemming Models, AI anthropology, and mass extinction events - how agents behave when left to govern themselves
As more and more systems make use of deep reasoning, humans are beginning to place more trust in them. But where there are multiple agents operating in adjacent areas, errors may often begin to occur. This is because agents modify their environment which then causes other agents to behave differently.
Software engineers have already seen this happen in a code base when code is modified. The behaviour change is prompted by the cultural artifacts that agents leave behind for later generations of agents. Against that context, Cognizant began a scientific experiment by setting up a platform to get a better understanding of agentic behavior over time.
As more enterprises are trusting agents to operate autonomously, Cognizant made the decision to provide full public access to this experimental platform called TerraLingua. The results are intriguing.
A petri dish to observe agentic behavior
As pitched by Babak Hodjat, Cognizant’s Chief AI Officer,...
Copyright of this story solely belongs to diginomica.com. To see the full text click HERE
